Statistical Learning Project - Supervised Learning
Read Dataset
ds = read.csv("AB_NYC_2019.csv")
head(ds)## id name host_id host_name
## 1 2539 Clean & quiet apt home by the park 2787 John
## 2 2595 Skylit Midtown Castle 2845 Jennifer
## 3 3647 THE VILLAGE OF HARLEM....NEW YORK ! 4632 Elisabeth
## 4 3831 Cozy Entire Floor of Brownstone 4869 LisaRoxanne
## 5 5022 Entire Apt: Spacious Studio/Loft by central park 7192 Laura
## 6 5099 Large Cozy 1 BR Apartment In Midtown East 7322 Chris
## neighbourhood_group neighbourhood latitude longitude room_type price
## 1 Brooklyn Kensington 40.64749 -73.97237 Private room 149
## 2 Manhattan Midtown 40.75362 -73.98377 Entire home/apt 225
## 3 Manhattan Harlem 40.80902 -73.94190 Private room 150
## 4 Brooklyn Clinton Hill 40.68514 -73.95976 Entire home/apt 89
## 5 Manhattan East Harlem 40.79851 -73.94399 Entire home/apt 80
## 6 Manhattan Murray Hill 40.74767 -73.97500 Entire home/apt 200
## minimum_nights number_of_reviews last_review reviews_per_month
## 1 1 9 2018-10-19 0.21
## 2 1 45 2019-05-21 0.38
## 3 3 0 NA
## 4 1 270 2019-07-05 4.64
## 5 10 9 2018-11-19 0.10
## 6 3 74 2019-06-22 0.59
## calculated_host_listings_count availability_365
## 1 6 365
## 2 2 355
## 3 1 365
## 4 1 194
## 5 1 0
## 6 1 129Data cleaning
Check for NA and NULL values
#Check for NA
apply(ds,2,function(x) sum(is.na(x)))## id name
## 0 0
## host_id host_name
## 0 0
## neighbourhood_group neighbourhood
## 0 0
## latitude longitude
## 0 0
## room_type price
## 0 0
## minimum_nights number_of_reviews
## 0 0
## last_review reviews_per_month
## 0 10052
## calculated_host_listings_count availability_365
## 0 0# NOTES
# Remove NA, empty
#
#
#
#Normalisation and selection of the variables
normalize <- function(x) {
return ((x - min(x)) / (max(x) - min(x)))
}
clean_data = function(ds)
{
ds = select (ds,-c(host_id, id, host_name, name,minimum_nights,number_of_reviews,neighbourhood,last_review,availability_365,
reviews_per_month,calculated_host_listings_count))
numerical = c("price","longitude", "latitude")
categorical = c("neighbourhood_group")
ds[numerical] = scale(ds[numerical])
ds$neighbourhood_group = factor(ds$neighbourhood_group,
level= c("Brooklyn","Manhattan",
"Queens","Staten Island", "Bronx"),
labels=c(1,2,3,4,5))
ds$room_type = factor(ds$room_type,
level= c("Private room","Entire home/apt","Shared room"),
labels=c(1,2,3))
return(ds)
}
#ggdraw() +
# draw_image("New_York_City_.png") +
# draw_plot(myplot)
dataset = clean_data(ds)
head(dataset)## neighbourhood_group latitude longitude room_type price
## 1 1 -1.4938339 -0.4376476 1 -0.01549291
## 2 2 0.4524314 -0.6846321 2 0.30097047
## 3 2 1.4683845 0.2224944 1 -0.01132892
## 4 1 -0.8033893 -0.1644481 2 -0.26533242
## 5 2 1.2756468 0.1772139 2 -0.30280835
## 6 2 0.3433173 -0.4946274 2 0.19687067Split data into train and test sets
library(caTools)## Warning: package 'caTools' was built under R version 3.6.3library(caret)## Warning: package 'caret' was built under R version 3.6.3## Loading required package: latticedata_clean = dataset
sample = sample.split(data_clean, SplitRatio = .75)
train = subset(data_clean, sample == TRUE)
test = subset(data_clean, sample == FALSE)
print("Initial data shape")## [1] "Initial data shape"print(dim(data_clean))## [1] 48895 5print("Train shape")## [1] "Train shape"print(dim(train))## [1] 29337 5print("Test shape")## [1] "Test shape"print(dim(test))## [1] 19558 5RANDOM FOREST
https://www.guru99.com/r-random-forest-tutorial.html
GOOD VIDEO FOR PARAMETER OPTIMISATION https://www.youtube.com/watch?v=6EXPYzbfLCE
TUNING https://uc-r.github.io/random_forests
library(randomForest)## Warning: package 'randomForest' was built under R version 3.6.3## randomForest 4.6-14## Type rfNews() to see new features/changes/bug fixes.##
## Attaching package: 'randomForest'## The following object is masked from 'package:dplyr':
##
## combine## The following object is masked from 'package:ggplot2':
##
## marginrf <- randomForest(
neighbourhood_group ~ . ,
data=train,
# importance = TRUE
)
rf##
## Call:
## randomForest(formula = neighbourhood_group ~ ., data = train, )
## Type of random forest: classification
## Number of trees: 500
## No. of variables tried at each split: 2
##
## OOB estimate of error rate: 0.13%
## Confusion matrix:
## 1 2 3 4 5 class.error
## 1 12083 1 12 0 0 0.0010747354
## 2 0 12947 0 0 2 0.0001544521
## 3 16 2 3387 0 0 0.0052863436
## 4 0 0 0 239 0 0.0000000000
## 5 0 4 0 0 644 0.0061728395#https://www.r-bloggers.com/how-to-implement-random-forests-in-r/
varImpPlot(rf)
pred = predict(rf, test)
head(pred)## 2 4 7 9 12 14
## 2 1 1 2 2 2
## Levels: 1 2 3 4 5plot(pred)
table(pred,test$neighbourhood_group)##
## pred 1 2 3 4 5
## 1 8005 2 7 0 0
## 2 1 8710 0 0 4
## 3 2 0 2253 0 0
## 4 0 0 0 134 0
## 5 0 0 1 0 439confusionMatrix(pred, test$neighbourhood_group)## Confusion Matrix and Statistics
##
## Reference
## Prediction 1 2 3 4 5
## 1 8005 2 7 0 0
## 2 1 8710 0 0 4
## 3 2 0 2253 0 0
## 4 0 0 0 134 0
## 5 0 0 1 0 439
##
## Overall Statistics
##
## Accuracy : 0.9991
## 95% CI : (0.9986, 0.9995)
## No Information Rate : 0.4454
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9986
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: 1 Class: 2 Class: 3 Class: 4 Class: 5
## Sensitivity 0.9996 0.9998 0.9965 1.000000 0.99097
## Specificity 0.9992 0.9995 0.9999 1.000000 0.99995
## Pos Pred Value 0.9989 0.9994 0.9991 1.000000 0.99773
## Neg Pred Value 0.9997 0.9998 0.9995 1.000000 0.99979
## Prevalence 0.4094 0.4454 0.1156 0.006851 0.02265
## Detection Rate 0.4093 0.4453 0.1152 0.006851 0.02245
## Detection Prevalence 0.4098 0.4456 0.1153 0.006851 0.02250
## Balanced Accuracy 0.9994 0.9997 0.9982 1.000000 0.99546rf2 <- randomForest(
price ~ . ,
data=train,
# importance = TRUE
)
rf2##
## Call:
## randomForest(formula = price ~ ., data = train, )
## Type of random forest: regression
## Number of trees: 500
## No. of variables tried at each split: 1
##
## Mean of squared residuals: 0.8848527
## % Var explained: 9.53varImpPlot(rf2)
par(mfrow=c(1,2))
plot(test$price,predict(rf2,test),col='red',main='Real vs predicted RF',pch=18,cex=0.7)
abline(0,1,lwd=2)
legend('bottomright',legend='NN',pch=18,col='red', bty='n')
library(randomForest)
library(e1071)## Warning: package 'e1071' was built under R version 3.6.3trControl <- trainControl(method = "cv",
number = 10,
search = "grid",
allowParallel = TRUE
)
trControl$method## [1] "cv"trControl$number## [1] 10Possible multi core tuning
registerDoFuture() plan(multiprocess, workers = availableCores() - 1)
library(doParallel) cl <- makeCluster(detectCores()) registerDoParallel(cl) #### machine learning code goes in here stopCluster(cl)
Calculate computational Time:
start.time <- Sys.time() …Relevent codes… end.time <- Sys.time() time.taken <- end.time - start.time time.taken
library(doParallel)## Warning: package 'doParallel' was built under R version 3.6.3## Loading required package: foreach## Warning: package 'foreach' was built under R version 3.6.3## Loading required package: iterators## Warning: package 'iterators' was built under R version 3.6.3## Loading required package: parallellibrary(doFuture)## Warning: package 'doFuture' was built under R version 3.6.3## Loading required package: globals## Loading required package: future##
## Attaching package: 'future'## The following object is masked from 'package:caret':
##
## clusterstart.time <- Sys.time()
#ptm <- proc.time()
rf_default <- caret::train(
neighbourhood_group ~ . ,
data=train,
method = "rf",
metric="Accuracy",
trControl = trControl,
num.threads = availableCores() # <- This one
)
end.time <- Sys.time()
time.taken <- end.time - start.time
print("-- Time: -- ")## [1] "-- Time: -- "time.taken## Time difference of 5.188274 minsprint("")## [1] ""print("-- RANDOM FOREST -- ")## [1] "-- RANDOM FOREST -- "#proc.time() - ptm
print(rf_default)## Random Forest
##
## 29337 samples
## 4 predictor
## 5 classes: '1', '2', '3', '4', '5'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 26402, 26402, 26405, 26405, 26403, 26403, ...
## Resampling results across tuning parameters:
##
## mtry Accuracy Kappa
## 2 0.9977504 0.9963760
## 3 0.9986025 0.9977494
## 5 0.9983979 0.9974200
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was mtry = 3.#ptm <- proc.time()
start.time <- Sys.time()
rf_default <- caret::train(
neighbourhood_group ~ . ,
data=train,
method = "rf",
metric="Accuracy",
trControl = trControl,
)
#proc.time() - ptm
end.time <- Sys.time()
time.taken <- end.time - start.time
print("-- Time: -- ")## [1] "-- Time: -- "time.taken## Time difference of 5.623895 minsprint("")## [1] ""print(rf_default)## Random Forest
##
## 29337 samples
## 4 predictor
## 5 classes: '1', '2', '3', '4', '5'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 26402, 26403, 26403, 26404, 26402, 26403, ...
## Resampling results across tuning parameters:
##
## mtry Accuracy Kappa
## 2 0.9979207 0.9966505
## 3 0.9988070 0.9980790
## 5 0.9986706 0.9978591
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was mtry = 3.start.time <- Sys.time()
set.seed(1234)
tuneGrid <- expand.grid(.mtry = c(1: 10))
rf_mtry <- caret::train(
neighbourhood_group~.,
data = train,
method = "rf",
metric = "Accuracy",
tuneGrid = tuneGrid,
trControl = trControl,
importance = TRUE,
nodesize = 14,
ntree = 300,
num.threads = availableCores()-1
)## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid range
## Warning in randomForest.default(x, y, mtry = param$mtry, ...): invalid mtry:
## reset to within valid rangeend.time <- Sys.time()
time.taken <- end.time - start.time
print("-- Time: -- ")## [1] "-- Time: -- "time.taken## Time difference of 10.68638 minsprint("")## [1] ""print(rf_mtry)## Random Forest
##
## 29337 samples
## 4 predictor
## 5 classes: '1', '2', '3', '4', '5'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 26404, 26403, 26402, 26403, 26403, 26404, ...
## Resampling results across tuning parameters:
##
## mtry Accuracy Kappa
## 1 0.9192489 0.8663930
## 2 0.9974435 0.9958825
## 3 0.9985684 0.9976949
## 4 0.9985002 0.9975854
## 5 0.9980911 0.9969263
## 6 0.9980911 0.9969265
## 7 0.9980911 0.9969263
## 8 0.9980571 0.9968714
## 9 0.9980230 0.9968167
## 10 0.9981593 0.9970361
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was mtry = 3.best_mtry <- rf_mtry$bestTune$mtry
cat("Best mtry value:",best_mtry)## Best mtry value: 3cat("\nMax accuracy mtry:", max(rf_mtry$results$Accuracy))##
## Max accuracy mtry: 0.9985684start.time <- Sys.time()
store_maxnode <- list()
tuneGrid <- expand.grid(.mtry = best_mtry)
for (maxnodes in c(5: 15)) {
set.seed(1234)
rf_maxnode <- train(neighbourhood_group~.,
data = train,
method = "rf",
metric = "Accuracy",
tuneGrid = tuneGrid,
trControl = trControl,
importance = TRUE,
nodesize = 14,
maxnodes = maxnodes,
ntree = 300,
)
current_iteration <- toString(maxnodes)
store_maxnode[[current_iteration]] <- rf_maxnode
}
end.time <- Sys.time()
time.taken <- end.time - start.time
print("-- Time: -- ")## [1] "-- Time: -- "time.taken## Time difference of 7.612453 minsprint("")## [1] ""results_mtry <- resamples(store_maxnode)
summary(results_mtry)##
## Call:
## summary.resamples(object = results_mtry)
##
## Models: 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
## Number of resamples: 10
##
## Accuracy
## Min. 1st Qu. Median Mean 3rd Qu. Max. NA's
## 5 0.8649847 0.8693425 0.8774922 0.8759245 0.8823079 0.8837368 0
## 6 0.8653256 0.8696834 0.8786850 0.8765380 0.8822626 0.8847596 0
## 7 0.8660075 0.8702284 0.8797072 0.8769128 0.8827139 0.8844187 0
## 8 0.8677122 0.8721134 0.8810301 0.8827754 0.8864081 0.9048756 0
## 9 0.8700989 0.8944766 0.9006476 0.8986593 0.9052407 0.9155026 0
## 10 0.9072938 0.9092459 0.9139543 0.9167942 0.9213263 0.9321976 0
## 11 0.9004432 0.9099121 0.9205728 0.9185315 0.9255470 0.9328333 0
## 12 0.9093079 0.9178385 0.9246888 0.9233723 0.9309640 0.9321976 0
## 13 0.9233129 0.9307024 0.9340605 0.9338720 0.9383150 0.9444255 0
## 14 0.9287419 0.9352253 0.9398434 0.9398368 0.9434270 0.9532901 0
## 15 0.9386294 0.9430811 0.9475109 0.9491422 0.9545801 0.9628367 0
##
## Kappa
## Min. 1st Qu. Median Mean 3rd Qu. Max. NA's
## 5 0.7793972 0.7864172 0.7996329 0.7971574 0.8076095 0.8098218 0
## 6 0.7799906 0.7868578 0.8016329 0.7980972 0.8071669 0.8117301 0
## 7 0.7811803 0.7881542 0.8033289 0.7989014 0.8083707 0.8113753 0
## 8 0.7838301 0.7910356 0.8055847 0.8085243 0.8146300 0.8450767 0
## 9 0.7876248 0.8273720 0.8376029 0.8343309 0.8453941 0.8618689 0
## 10 0.8483833 0.8515134 0.8592176 0.8639096 0.8713883 0.8890951 0
## 11 0.8369987 0.8526091 0.8701376 0.8667766 0.8782680 0.8904073 0
## 12 0.8516275 0.8656820 0.8769131 0.8747760 0.8872923 0.8891170 0
## 13 0.8746033 0.8868985 0.8923119 0.8920415 0.8994440 0.9094351 0
## 14 0.8836125 0.8943161 0.9016622 0.9017820 0.9077138 0.9238400 0
## 15 0.8998846 0.9071320 0.9141669 0.9169273 0.9257702 0.9392863 0start.time <- Sys.time()
for (maxnodes in c(15 : 30)) {
set.seed(1234)
rf_maxnode <- train(neighbourhood_group~.,
data = train,
method = "rf",
metric = "Accuracy",
tuneGrid = tuneGrid,
trControl = trControl,
importance = TRUE,
nodesize = 14,
maxnodes = maxnodes,
ntree = 300)
current_iteration <- toString(maxnodes)
store_maxnode[[current_iteration]] <- rf_maxnode
}
end.time <- Sys.time()
time.taken <- end.time - start.time
print("-- Time: -- ")## [1] "-- Time: -- "time.taken## Time difference of 14.05081 minsprint("")## [1] ""results_mtry <- resamples(store_maxnode)
summary(results_mtry)##
## Call:
## summary.resamples(object = results_mtry)
##
## Models: 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30
## Number of resamples: 10
##
## Accuracy
## Min. 1st Qu. Median Mean 3rd Qu. Max. NA's
## 5 0.8649847 0.8693425 0.8774922 0.8759245 0.8823079 0.8837368 0
## 6 0.8653256 0.8696834 0.8786850 0.8765380 0.8822626 0.8847596 0
## 7 0.8660075 0.8702284 0.8797072 0.8769128 0.8827139 0.8844187 0
## 8 0.8677122 0.8721134 0.8810301 0.8827754 0.8864081 0.9048756 0
## 9 0.8700989 0.8944766 0.9006476 0.8986593 0.9052407 0.9155026 0
## 10 0.9072938 0.9092459 0.9139543 0.9167942 0.9213263 0.9321976 0
## 11 0.9004432 0.9099121 0.9205728 0.9185315 0.9255470 0.9328333 0
## 12 0.9093079 0.9178385 0.9246888 0.9233723 0.9309640 0.9321976 0
## 13 0.9233129 0.9307024 0.9340605 0.9338720 0.9383150 0.9444255 0
## 14 0.9287419 0.9352253 0.9398434 0.9398368 0.9434270 0.9532901 0
## 15 0.9386294 0.9430811 0.9475109 0.9491422 0.9545801 0.9628367 0
## 16 0.9458078 0.9495397 0.9529734 0.9542213 0.9597878 0.9635186 0
## 17 0.9515854 0.9556843 0.9587594 0.9593693 0.9642095 0.9682919 0
## 18 0.9519264 0.9562883 0.9613024 0.9607653 0.9656668 0.9696763 0
## 19 0.9498807 0.9575665 0.9631776 0.9628796 0.9673624 0.9741056 0
## 20 0.9567144 0.9601304 0.9708590 0.9679253 0.9734061 0.9778459 0
## 21 0.9706785 0.9719570 0.9759713 0.9751506 0.9770810 0.9809069 0
## 22 0.9723926 0.9750297 0.9768155 0.9770937 0.9791259 0.9819359 0
## 23 0.9734151 0.9760484 0.9769938 0.9776731 0.9804040 0.9819359 0
## 24 0.9737470 0.9762189 0.9780164 0.9780140 0.9799779 0.9819359 0
## 25 0.9751193 0.9779236 0.9786982 0.9795819 0.9827000 0.9839809 0
## 26 0.9737470 0.9786091 0.9798841 0.9796161 0.9822813 0.9843164 0
## 27 0.9758010 0.9796284 0.9812546 0.9811841 0.9837267 0.9849983 0
## 28 0.9771643 0.9788722 0.9809069 0.9810819 0.9834697 0.9853392 0
## 29 0.9774974 0.9800648 0.9809069 0.9817295 0.9841527 0.9863621 0
## 30 0.9781868 0.9799693 0.9819363 0.9821385 0.9846601 0.9856851 0
##
## Kappa
## Min. 1st Qu. Median Mean 3rd Qu. Max. NA's
## 5 0.7793972 0.7864172 0.7996329 0.7971574 0.8076095 0.8098218 0
## 6 0.7799906 0.7868578 0.8016329 0.7980972 0.8071669 0.8117301 0
## 7 0.7811803 0.7881542 0.8033289 0.7989014 0.8083707 0.8113753 0
## 8 0.7838301 0.7910356 0.8055847 0.8085243 0.8146300 0.8450767 0
## 9 0.7876248 0.8273720 0.8376029 0.8343309 0.8453941 0.8618689 0
## 10 0.8483833 0.8515134 0.8592176 0.8639096 0.8713883 0.8890951 0
## 11 0.8369987 0.8526091 0.8701376 0.8667766 0.8782680 0.8904073 0
## 12 0.8516275 0.8656820 0.8769131 0.8747760 0.8872923 0.8891170 0
## 13 0.8746033 0.8868985 0.8923119 0.8920415 0.8994440 0.9094351 0
## 14 0.8836125 0.8943161 0.9016622 0.9017820 0.9077138 0.9238400 0
## 15 0.8998846 0.9071320 0.9141669 0.9169273 0.9257702 0.9392863 0
## 16 0.9117547 0.9178268 0.9235396 0.9255196 0.9345687 0.9407029 0
## 17 0.9212544 0.9279546 0.9330828 0.9339460 0.9417676 0.9485093 0
## 18 0.9218246 0.9290189 0.9370455 0.9362336 0.9441766 0.9506521 0
## 19 0.9185262 0.9311699 0.9401894 0.9397016 0.9469434 0.9579223 0
## 20 0.9296463 0.9351900 0.9526127 0.9479116 0.9568140 0.9640983 0
## 21 0.9523150 0.9544444 0.9609792 0.9596524 0.9628454 0.9689943 0
## 22 0.9551685 0.9594814 0.9623503 0.9628126 0.9661463 0.9707219 0
## 23 0.9567988 0.9610981 0.9626597 0.9637459 0.9682097 0.9706991 0
## 24 0.9573366 0.9613930 0.9642862 0.9642935 0.9675180 0.9707000 0
## 25 0.9595853 0.9641321 0.9653823 0.9668399 0.9719168 0.9740505 0
## 26 0.9573428 0.9652395 0.9673369 0.9669018 0.9712556 0.9745665 0
## 27 0.9607108 0.9668990 0.9695829 0.9694608 0.9736205 0.9756806 0
## 28 0.9628847 0.9656645 0.9690203 0.9692872 0.9732049 0.9762372 0
## 29 0.9634154 0.9676231 0.9690040 0.9703438 0.9743209 0.9779104 0
## 30 0.9645500 0.9674709 0.9706882 0.9710099 0.9751228 0.9768119 0Result: 30
Step 4) Search the best ntrees
Now that you have the best value of mtry and maxnode, you can tune the number of trees. The method is exactly the same as maxnode.
start.time <- Sys.time()
store_maxtrees <- list()
for (ntree in c(250, 300, 350, 400, 450, 500, 550, 600, 800, 1000, 2000)) {
set.seed(5678)
rf_maxtrees <- train(neighbourhood_group~.,
data = train,
method = "rf",
metric = "Accuracy",
tuneGrid = tuneGrid,
trControl = trControl,
importance = TRUE,
nodesize = 14,
maxnodes = 24,
ntree = ntree)
key <- toString(ntree)
store_maxtrees[[key]] <- rf_maxtrees
}
end.time <- Sys.time()
time.taken <- end.time - start.time
print("-- Time: -- ")## [1] "-- Time: -- "time.taken## Time difference of 20.17586 minsprint("") ## [1] ""results_tree <- resamples(store_maxtrees)
summary(results_tree)##
## Call:
## summary.resamples(object = results_tree)
##
## Models: 250, 300, 350, 400, 450, 500, 550, 600, 800, 1000, 2000
## Number of resamples: 10
##
## Accuracy
## Min. 1st Qu. Median Mean 3rd Qu. Max. NA's
## 250 0.9754518 0.9769126 0.9783535 0.9785596 0.9797954 0.9822768 0
## 300 0.9761337 0.9773386 0.9781831 0.9788663 0.9799660 0.9826176 0
## 350 0.9754518 0.9775827 0.9778497 0.9788663 0.9802284 0.9826176 0
## 400 0.9754518 0.9769128 0.9781905 0.9787640 0.9801364 0.9826176 0
## 450 0.9757927 0.9775942 0.9780201 0.9789003 0.9802284 0.9826176 0
## 500 0.9757927 0.9775128 0.9778422 0.9787299 0.9798022 0.9826176 0
## 550 0.9754518 0.9771604 0.9776794 0.9786617 0.9798022 0.9829584 0
## 600 0.9757927 0.9772456 0.9778498 0.9786958 0.9797954 0.9826176 0
## 800 0.9757927 0.9775942 0.9778497 0.9787980 0.9798022 0.9826176 0
## 1000 0.9757927 0.9775013 0.9776794 0.9787299 0.9797954 0.9826176 0
## 2000 0.9754518 0.9772534 0.9780201 0.9786958 0.9798022 0.9822768 0
##
## Kappa
## Min. 1st Qu. Median Mean 3rd Qu. Max. NA's
## 250 0.9600825 0.9624691 0.9648381 0.9651800 0.9672193 0.9712526 0
## 300 0.9611893 0.9631733 0.9645575 0.9656789 0.9674895 0.9718120 0
## 350 0.9600832 0.9635485 0.9640089 0.9656738 0.9679314 0.9718105 0
## 400 0.9600667 0.9624696 0.9645621 0.9655031 0.9677586 0.9718105 0
## 450 0.9606278 0.9635856 0.9642846 0.9657281 0.9679314 0.9718100 0
## 500 0.9606278 0.9634492 0.9640085 0.9654539 0.9672477 0.9718100 0
## 550 0.9600667 0.9629186 0.9637187 0.9653449 0.9672477 0.9723673 0
## 600 0.9606278 0.9630503 0.9640081 0.9654002 0.9672160 0.9718100 0
## 800 0.9606278 0.9635816 0.9640128 0.9655633 0.9672457 0.9718105 0
## 1000 0.9606278 0.9634492 0.9637324 0.9654556 0.9672160 0.9718105 0
## 2000 0.9600825 0.9630424 0.9642846 0.9654014 0.9672509 0.9712526 0Step 5) Evaluate the model
fit_rf <- train(neighbourhood_group~.,
train,
method = "rf",
metric = "Accuracy",
tuneGrid = tuneGrid,
trControl = trControl,
importance = TRUE,
nodesize = 14,
ntree = 1000,
maxnodes = 30)
prediction <-predict(fit_rf, test)
prediction_default <-predict(rf, test)
conf = confusionMatrix(prediction, test$neighbourhood_group)
conf_default = confusionMatrix(prediction_default, test$neighbourhood_group)
conf## Confusion Matrix and Statistics
##
## Reference
## Prediction 1 2 3 4 5
## 1 7999 5 124 0 0
## 2 5 8697 171 0 22
## 3 4 0 1966 0 0
## 4 0 0 0 134 0
## 5 0 10 0 0 421
##
## Overall Statistics
##
## Accuracy : 0.9826
## 95% CI : (0.9806, 0.9844)
## No Information Rate : 0.4454
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9717
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: 1 Class: 2 Class: 3 Class: 4 Class: 5
## Sensitivity 0.9989 0.9983 0.8695 1.000000 0.95034
## Specificity 0.9888 0.9817 0.9998 1.000000 0.99948
## Pos Pred Value 0.9841 0.9777 0.9980 1.000000 0.97680
## Neg Pred Value 0.9992 0.9986 0.9832 1.000000 0.99885
## Prevalence 0.4094 0.4454 0.1156 0.006851 0.02265
## Detection Rate 0.4090 0.4447 0.1005 0.006851 0.02153
## Detection Prevalence 0.4156 0.4548 0.1007 0.006851 0.02204
## Balanced Accuracy 0.9939 0.9900 0.9346 1.000000 0.97491conf_default## Confusion Matrix and Statistics
##
## Reference
## Prediction 1 2 3 4 5
## 1 8005 2 7 0 0
## 2 1 8710 0 0 4
## 3 2 0 2253 0 0
## 4 0 0 0 134 0
## 5 0 0 1 0 439
##
## Overall Statistics
##
## Accuracy : 0.9991
## 95% CI : (0.9986, 0.9995)
## No Information Rate : 0.4454
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9986
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: 1 Class: 2 Class: 3 Class: 4 Class: 5
## Sensitivity 0.9996 0.9998 0.9965 1.000000 0.99097
## Specificity 0.9992 0.9995 0.9999 1.000000 0.99995
## Pos Pred Value 0.9989 0.9994 0.9991 1.000000 0.99773
## Neg Pred Value 0.9997 0.9998 0.9995 1.000000 0.99979
## Prevalence 0.4094 0.4454 0.1156 0.006851 0.02265
## Detection Rate 0.4093 0.4453 0.1152 0.006851 0.02245
## Detection Prevalence 0.4098 0.4456 0.1153 0.006851 0.02250
## Balanced Accuracy 0.9994 0.9997 0.9982 1.000000 0.99546#Step 6) Visualize Result
#varImpPlot(fit_rf)RANGER RANDOM FOREST
library(ranger)## Warning: package 'ranger' was built under R version 3.6.3##
## Attaching package: 'ranger'## The following object is masked from 'package:randomForest':
##
## importancelibrary(tuneRanger)## Warning: package 'tuneRanger' was built under R version 3.6.3## Loading required package: mlrMBO## Warning: package 'mlrMBO' was built under R version 3.6.3## Loading required package: mlr## Warning: package 'mlr' was built under R version 3.6.3## Loading required package: ParamHelpers## Warning: package 'ParamHelpers' was built under R version 3.6.3## 'mlr' is in maintenance mode since July 2019. Future development
## efforts will go into its successor 'mlr3' (<https://mlr3.mlr-org.com>).##
## Attaching package: 'mlr'## The following object is masked from 'package:e1071':
##
## impute## The following object is masked from 'package:caret':
##
## train## Loading required package: smoof## Warning: package 'smoof' was built under R version 3.6.3## Loading required package: checkmate## Loading required package: lubridate## Warning: package 'lubridate' was built under R version 3.6.3##
## Attaching package: 'lubridate'## The following objects are masked from 'package:dplyr':
##
## intersect, setdiff, union## The following object is masked from 'package:cowplot':
##
## stamp## The following objects are masked from 'package:base':
##
## date, intersect, setdiff, union## Loading required package: lhs## Warning: package 'lhs' was built under R version 3.6.3ranger <- ranger( neighbourhood_group~ ., data = train, write.forest = TRUE, classification = T)
ranger_pred = predict(ranger, data = test)
tab = table(test$neighbourhood_group, predictions(ranger_pred))
plot(tab)
rangerReg <- ranger( price~ ., data = train, write.forest = TRUE, classification = F)
rangerReg## Ranger result
##
## Call:
## ranger(price ~ ., data = train, write.forest = TRUE, classification = F)
##
## Type: Regression
## Number of trees: 500
## Sample size: 29337
## Number of independent variables: 4
## Mtry: 2
## Target node size: 5
## Variable importance mode: none
## Splitrule: variance
## OOB prediction error (MSE): 0.8986566
## R squared (OOB): 0.08119909rangerReg_pred = predict(rangerReg, data = test)
rangerReg_pred## Ranger prediction
##
## Type: Regression
## Sample size: 19558
## Number of independent variables: 4tab = table(test$price, predictions(rangerReg_pred))
#tab
library(plotly)## Warning: package 'plotly' was built under R version 3.6.3##
## Attaching package: 'plotly'## The following object is masked from 'package:ggmap':
##
## wind## The following object is masked from 'package:ggplot2':
##
## last_plot## The following object is masked from 'package:stats':
##
## filter## The following object is masked from 'package:graphics':
##
## layoutplot_ly(x = test$price, y = predictions(rangerReg_pred))## No trace type specified:
## Based on info supplied, a 'scatter' trace seems appropriate.
## Read more about this trace type -> https://plot.ly/r/reference/#scatter## No scatter mode specifed:
## Setting the mode to markers
## Read more about this attribute -> https://plot.ly/r/reference/#scatter-modeRANGER TUNING
library(tuneRanger)
# https://github.com/PhilippPro/tuneRanger
# https://mlr.mlr-org.com/articles/tutorial/measures.html
task = makeRegrTask(data = train, target = "price")
estimateTimeTuneRanger(task, iters = 20, num.threads = 8, num.trees = 1000)## Approximated time for tuning: 4M 3Sres = tuneRanger(task, measure = list(mse), num.trees = 1000,
num.threads = 8, iters = 20, show.info = getOption("mlrMBO.show.info", TRUE))## Computing y column(s) for design. Not provided.## [mbo] 0: mtry=1; min.node.size=230; sample.fraction=0.777 : y = 0.885 : 3.0 secs : initdesign## [mbo] 0: mtry=1; min.node.size=933; sample.fraction=0.449 : y = 0.886 : 2.3 secs : initdesign## [mbo] 0: mtry=3; min.node.size=8; sample.fraction=0.749 : y = 0.979 : 27.0 secs : initdesign## [mbo] 0: mtry=2; min.node.size=19; sample.fraction=0.549 : y = 0.889 : 8.8 secs : initdesign## [mbo] 0: mtry=3; min.node.size=56; sample.fraction=0.654 : y = 0.911 : 17.3 secs : initdesign## [mbo] 0: mtry=2; min.node.size=1.1e+03; sample.fraction=0.233 : y = 0.885 : 2.4 secs : initdesign## [mbo] 0: mtry=4; min.node.size=129; sample.fraction=0.373 : y = 0.888 : 9.3 secs : initdesign## [mbo] 0: mtry=2; min.node.size=2.18e+03; sample.fraction=0.509 : y = 0.885 : 2.4 secs : initdesign## [mbo] 0: mtry=4; min.node.size=6; sample.fraction=0.809 : y = 1.11 : 36.1 secs : initdesign## [mbo] 0: mtry=1; min.node.size=2; sample.fraction=0.463 : y = 0.885 : 2.6 secs : initdesign## [mbo] 0: mtry=3; min.node.size=66; sample.fraction=0.223 : y = 0.878 : 5.0 secs : initdesign## [mbo] 0: mtry=3; min.node.size=338; sample.fraction=0.359 : y = 0.884 : 4.0 secs : initdesign## [mbo] 0: mtry=3; min.node.size=4; sample.fraction=0.334 : y = 0.91 : 11.6 secs : initdesign## [mbo] 0: mtry=1; min.node.size=11; sample.fraction=0.485 : y = 0.884 : 2.9 secs : initdesign## [mbo] 0: mtry=4; min.node.size=561; sample.fraction=0.7 : y = 0.902 : 7.9 secs : initdesign## [mbo] 0: mtry=4; min.node.size=5.02e+03; sample.fraction=0.598 : y = 0.894 : 2.3 secs : initdesign## [mbo] 0: mtry=2; min.node.size=282; sample.fraction=0.572 : y = 0.882 : 5.2 secs : initdesign## [mbo] 0: mtry=2; min.node.size=35; sample.fraction=0.717 : y = 0.891 : 8.6 secs : initdesign## [mbo] 0: mtry=3; min.node.size=32; sample.fraction=0.398 : y = 0.895 : 10.8 secs : initdesign## [mbo] 0: mtry=1; min.node.size=3.01e+03; sample.fraction=0.674 : y = 0.89 : 2.0 secs : initdesign## [mbo] 0: mtry=3; min.node.size=4; sample.fraction=0.635 : y = 0.971 : 26.8 secs : initdesign## [mbo] 0: mtry=4; min.node.size=16; sample.fraction=0.59 : y = 0.949 : 24.4 secs : initdesign## [mbo] 0: mtry=2; min.node.size=4e+03; sample.fraction=0.868 : y = 0.886 : 2.9 secs : initdesign## [mbo] 0: mtry=1; min.node.size=177; sample.fraction=0.839 : y = 0.885 : 3.1 secs : initdesign## [mbo] 0: mtry=2; min.node.size=6; sample.fraction=0.251 : y = 0.883 : 6.2 secs : initdesign## [mbo] 0: mtry=4; min.node.size=1.5e+03; sample.fraction=0.303 : y = 0.89 : 2.7 secs : initdesign## [mbo] 0: mtry=1; min.node.size=3; sample.fraction=0.799 : y = 0.885 : 3.3 secs : initdesign## [mbo] 0: mtry=4; min.node.size=82; sample.fraction=0.898 : y = 0.999 : 20.8 secs : initdesign## [mbo] 0: mtry=4; min.node.size=2; sample.fraction=0.275 : y = 0.916 : 14.1 secs : initdesign## [mbo] 0: mtry=1; min.node.size=759; sample.fraction=0.415 : y = 0.886 : 2.4 secs : initdesign## [mbo] 1: mtry=1; min.node.size=3; sample.fraction=0.201 : y = 0.885 : 3.0 secs : infill_cb## [mbo] 2: mtry=4; min.node.size=44; sample.fraction=0.212 : y = 0.883 : 8.5 secs : infill_cb## [mbo] 3: mtry=4; min.node.size=496; sample.fraction=0.514 : y = 0.895 : 6.4 secs : infill_cb## [mbo] 4: mtry=2; min.node.size=866; sample.fraction=0.719 : y = 0.883 : 4.2 secs : infill_cb## [mbo] 5: mtry=2; min.node.size=99; sample.fraction=0.293 : y = 0.879 : 4.6 secs : infill_cb## [mbo] 6: mtry=1; min.node.size=15; sample.fraction=0.703 : y = 0.885 : 3.2 secs : infill_cb## [mbo] 7: mtry=3; min.node.size=4.66e+03; sample.fraction=0.215 : y = 0.907 : 1.6 secs : infill_cb## [mbo] 8: mtry=2; min.node.size=29; sample.fraction=0.201 : y = 0.88 : 4.7 secs : infill_cb## [mbo] 9: mtry=3; min.node.size=194; sample.fraction=0.203 : y = 0.882 : 3.9 secs : infill_cb## [mbo] 10: mtry=1; min.node.size=2; sample.fraction=0.9 : y = 0.885 : 3.1 secs : infill_cb## [mbo] 11: mtry=1; min.node.size=16; sample.fraction=0.275 : y = 0.885 : 2.5 secs : infill_cb## [mbo] 12: mtry=1; min.node.size=24; sample.fraction=0.889 : y = 0.885 : 3.2 secs : infill_cb## [mbo] 13: mtry=1; min.node.size=5.34e+03; sample.fraction=0.304 : y = 0.921 : 1.9 secs : infill_cb## [mbo] 14: mtry=1; min.node.size=2.33e+03; sample.fraction=0.898 : y = 0.888 : 1.9 secs : infill_cb## [mbo] 15: mtry=2; min.node.size=525; sample.fraction=0.385 : y = 0.882 : 3.7 secs : infill_cb## [mbo] 16: mtry=1; min.node.size=98; sample.fraction=0.564 : y = 0.885 : 2.6 secs : infill_cb## [mbo] 17: mtry=3; min.node.size=24; sample.fraction=0.242 : y = 0.885 : 8.4 secs : infill_cb## [mbo] 18: mtry=2; min.node.size=69; sample.fraction=0.225 : y = 0.879 : 5.0 secs : infill_cb## [mbo] 19: mtry=1; min.node.size=2; sample.fraction=0.33 : y = 0.885 : 2.7 secs : infill_cb## [mbo] 20: mtry=3; min.node.size=97; sample.fraction=0.276 : y = 0.88 : 7.4 secs : infill_cbMean of best 5 % of the results
res## Recommended parameter settings:
## mtry min.node.size sample.fraction
## 1 2 78 0.2467632
## Results:
## mse exec.time
## 1 0.8787459 4.84Recommended parameter settings: mtry min.node.size sample.fraction 1 2 55 0.2136541 Results: mse exec.time 1 0.933998 2.756667
Model with the new tuned hyperparameters
res$model## Model for learner.id=regr.ranger; learner.class=regr.ranger
## Trained on: task.id = train; obs = 29337; features = 4
## Hyperparameters: num.threads=8,verbose=FALSE,respect.unordered.factors=order,mtry=2,min.node.size=78,sample.fraction=0.247,num.trees=1e+03,replace=FALSEtuned_rangerReg <- ranger( price~ ., data = train, write.forest = TRUE, classification = F, mtry= 2,
min.node.size = 55, sample.fraction = 0.214,num.trees = 1000, replace= FALSE)
tuned_rangerReg## Ranger result
##
## Call:
## ranger(price ~ ., data = train, write.forest = TRUE, classification = F, mtry = 2, min.node.size = 55, sample.fraction = 0.214, num.trees = 1000, replace = FALSE)
##
## Type: Regression
## Number of trees: 1000
## Sample size: 29337
## Number of independent variables: 4
## Mtry: 2
## Target node size: 55
## Variable importance mode: none
## Splitrule: variance
## OOB prediction error (MSE): 0.8787042
## R squared (OOB): 0.1015987tuned_rangerReg_pred = predict(tuned_rangerReg, data = test)
tuned_rangerReg_pred## Ranger prediction
##
## Type: Regression
## Sample size: 19558
## Number of independent variables: 4ROCK CURVE
cm = table(test[,14], pred)
library(caret) confusionMatrix(as.factor(pred), as.factor(test$price))
confusionMatrix( factor(pred, levels = 1:19558), factor(test$price, levels = 1:19558) )
library(ROCR)## Warning: package 'ROCR' was built under R version 3.6.3##
## Attaching package: 'ROCR'## The following object is masked from 'package:mlr':
##
## performance#ROCR
#roc_pred = predict(tuned_rangerReg,data = test, type="response")
#perf = prediction(roc_pred$predictions, test)
#==
#FIND OPTIMAL VALUE WITH MIN OUT OF BAG
#https://www.listendata.com/2014/11/random-forest-with-r.htmlDECISION TREE ——– TO BE DONE
library(rpart) library(caret) library(e1071)
dt = train( neighbourhood_group~ ., data = train,method = “rpart”) pred2 = predict(dt, data = test) table(pred2, test)
NEURAL NETWORKS ??
https://medium.com/@brscntyz/neural-network-in-r-e275302b6e44 https://datascienceplus.com/fitting-neural-network-in-r/ https://www.kdnuggets.com/2016/08/begineers-guide-neural-networks-r.html/2
library(ISLR)## Warning: package 'ISLR' was built under R version 3.6.3library(tidyverse)## -- Attaching packages --------------------------------------- tidyverse 1.3.0 --## v tibble 3.0.1 v stringr 1.4.0
## v readr 1.3.1 v forcats 0.5.0
## v purrr 0.3.4## Warning: package 'tibble' was built under R version 3.6.3## Warning: package 'stringr' was built under R version 3.6.3## Warning: package 'forcats' was built under R version 3.6.3## -- Conflicts ------------------------------------------ tidyverse_conflicts() --
## x purrr::accumulate() masks foreach::accumulate()
## x lubridate::as.difftime() masks base::as.difftime()
## x randomForest::combine() masks dplyr::combine()
## x lubridate::date() masks base::date()
## x plotly::filter() masks dplyr::filter(), stats::filter()
## x lubridate::intersect() masks base::intersect()
## x dplyr::lag() masks stats::lag()
## x purrr::lift() masks caret::lift()
## x randomForest::margin() masks ggplot2::margin()
## x lubridate::setdiff() masks base::setdiff()
## x lubridate::stamp() masks cowplot::stamp()
## x lubridate::union() masks base::union()
## x purrr::when() masks foreach::when()library("keras")## Warning: package 'keras' was built under R version 3.6.3##
## Attaching package: 'keras'## The following object is masked _by_ '.GlobalEnv':
##
## normalize## The following object is masked from 'package:future':
##
## %<-%library(neuralnet)## Warning: package 'neuralnet' was built under R version 3.6.3##
## Attaching package: 'neuralnet'## The following object is masked from 'package:ROCR':
##
## prediction## The following object is masked from 'package:dplyr':
##
## computelibrary(Hmisc)## Loading required package: survival##
## Attaching package: 'survival'## The following object is masked from 'package:future':
##
## cluster## The following object is masked from 'package:caret':
##
## cluster## Loading required package: Formula##
## Attaching package: 'Hmisc'## The following object is masked from 'package:plotly':
##
## subplot## The following object is masked from 'package:mlr':
##
## impute## The following object is masked from 'package:e1071':
##
## impute## The following objects are masked from 'package:dplyr':
##
## src, summarize## The following objects are masked from 'package:base':
##
## format.pval, unitsm <- model.matrix(
~price+neighbourhood_group+room_type+longitude+latitude,
data = train
)
m_test <- model.matrix(
~price+neighbourhood_group+room_type+longitude+latitude,
data = test
)
head(m)## (Intercept) price neighbourhood_group2 neighbourhood_group3
## 1 1 -0.01549291 0 0
## 3 1 -0.01132892 1 0
## 5 1 -0.30280835 1 0
## 6 1 0.19687067 1 0
## 8 1 -0.30697234 1 0
## 10 1 -0.01132892 1 0
## neighbourhood_group4 neighbourhood_group5 room_type2 room_type3 longitude
## 1 0 0 0 0 -0.4376476
## 3 0 0 0 0 0.2224944
## 5 0 0 1 0 0.1772139
## 6 0 0 1 0 -0.4946274
## 8 0 0 0 0 -0.7097639
## 10 0 0 1 0 -0.8276232
## latitude
## 1 -1.4938339
## 3 1.4683845
## 5 1.2756468
## 6 0.3433173
## 8 0.6591063
## 10 -0.2844097#nn=neuralnet(price~ neighbourhood_group2+ neighbourhood_group3 +neighbourhood_group4+ neighbourhood_group5+ room_type2+ room_type3+longitude+latitude,data=m, hidden=10,act.fct = "logistic",
# linear.output = TRUE,stepmax=10^5,threshold = 0.01)Also you can change your hidden layers by specifiying with numbers in vector like this
#nn=neuralnet( price~ neighbourhood_group2+ neighbourhood_group3 +neighbourhood_group4+ neighbourhood_group5+ room_type2+ room_type3+longitude+latitude,data=m, hidden=c(7,6,5),act.fct = "logistic",
# linear.output = TRUE,stepmax=10^5,threshold = 0.01)
#hidden=c(7,6,5)Then, prediction and calculation of error comes. I calculate the error with Root mean error method.
nn_pred=compute(nn,test[,1:13]) nn_pred\(net.resultRMSE <- function(actual,predicted) { return(sqrt(sum(actual^2-predicted^2)/length(actual))) } summary(nn_pred) nn_pred <- is.numeric(nn_pred) RMSE(test\)price,nn_pred)
plot(test\(price,nn_pred\)net.result)
pr.nn_ <- nn_pred\(net.result*(max(dataset\)price)-min(dataset\(price))+min(dataset\)price)
LINEAR REGRESSION
https://datascienceplus.com/fitting-neural-network-in-r/
lm.fit <- glm(price~., data=train)
summary(lm.fit)##
## Call:
## glm(formula = price ~ ., data = train)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.009 -0.254 -0.102 0.044 41.017
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.361387 0.012348 -29.267 < 2e-16 ***
## neighbourhood_group2 0.271929 0.019299 14.091 < 2e-16 ***
## neighbourhood_group3 0.112169 0.023815 4.710 2.49e-06 ***
## neighbourhood_group4 -0.463934 0.068425 -6.780 1.22e-11 ***
## neighbourhood_group5 0.189675 0.046677 4.064 4.85e-05 ***
## latitude -0.054288 0.009057 -5.994 2.07e-09 ***
## longitude -0.080103 0.008788 -9.115 < 2e-16 ***
## room_type2 0.440940 0.011498 38.348 < 2e-16 ***
## room_type3 -0.091403 0.036869 -2.479 0.0132 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for gaussian family taken to be 0.893665)
##
## Null deviance: 28693 on 29336 degrees of freedom
## Residual deviance: 26209 on 29328 degrees of freedom
## AIC: 79968
##
## Number of Fisher Scoring iterations: 2pr.lm <- predict(lm.fit,test)
MSE.lm <- sum((pr.lm - test$price)^2)/nrow(test)par(mfrow=c(1,2))
#plot(test$price,nn_pred$net.result,col='red',main='Real vs predicted NN',pch=18,cex=0.7)
#abline(0,1,lwd=2)
#legend('bottomright',legend='NN',pch=18,col='red', bty='n')
plot(test$price,pr.lm,col='blue',main='Real vs predicted Linear Regression',pch=18, cex=0.7)
abline(0,1,lwd=2)
legend('bottomright',legend='LM',pch=18,col='blue', bty='n', cex=.95)